Introduction

Selective serotonin reuptake inhibitors (SSRIs) are a class of medications commonly prescribed to treat depression, anxiety, and other mental health conditions. They work by increasing the availability of serotonin, a neurotransmitter that helps regulate mood. Due to their broad application and generally favourable side effect profile, SSRIs are a key indicator of prescribing patterns linked to mental health needs within populations (Madsen et al., 2015). Mental health, however, is not static and can be influenced by various external factors, including environmental changes. Seasonal fluctuations in mood and mental health, often referred to as Seasonal Affective Disorder (SAD), are well-documented in medical literature (Lam et al., 2006). SAD is typically triggered by reduced daylight in the autumn and winter months, leading to symptoms such as low mood, anxiety, and feelings of hopelessness. While not everyone who experiences seasonal mood shifts meets the criteria for SAD, many individuals seek medical intervention, often in the form of antidepressant prescriptions (Harmatz et al., 2000).

In Scotland, where seasonal variations are pronounced—such as extended periods of darkness in winter and unpredictable weather—these environmental factors may exacerbate mental health challenges. This could contribute to increased SSRI prescriptions during certain times of the year. Previous studies have shown that seasonal trends in antidepressant prescribing are influenced not only by environmental factors but also by societal pressures and healthcare access(Jack et al., 2023).

This report examines the seasonal trends in SSRI prescriptions across Scotland’s three major health boards, which represent the largest and primarily urban populations. Using the Public Health Scotland prescribing data, available here: https://www.opendata.nhs.scot/dataset/prescriptions-in-the-community

The analysis aims to identify whether SSRI prescriptions show distinct seasonal patterns. While this analysis focuses solely on SSRI prescription data and does not include specific conditions for which they are prescribed. Any suggested causes for these trends are speculative, as the data is limited to prescription counts rather than diagnoses or underlying factors.

By presenting these patterns, this report aims to contribute to a better understanding of SSRI use across Scotland’s major health boards and highlight areas for further investigation into the broader factors that may drive these trends.

Load Necessary Libraries

#load necessary libraries 
library(tidyverse) 
library(janitor) 
library(gt)
library(here)
library(plotly)
# Function to process prescription data from multiple CSV files
process_prescription_data <- function(file_paths, health_boards) {
  # Step 1: Load and combine the files
  processed_data <- map_dfr(file_paths, ~ {
    message("Processing file: ", .x)  # Debugging log
    # Read the file, select relevant columns and filter relevant rows
    read_csv(.x, col_types = cols()) %>%
      select(HBT2014, BNFItemDescription, PaidQuantity, PaidDateMonth) %>%  
      filter(str_detect(BNFItemDescription, "^(SERTRALINE|PAROXETINE|FLUOXETINE|CITALOPRAM|ESCITALOPRAM)") & 
               HBT2014 %in% health_boards)})
  # Step 2: Cleaning and transforming the data by using mutate
  full_prescribed_data <- processed_data %>%
    mutate( # Retaining the PaidDateMonth in YYYYMMDD format
      PaidDateMonth = sprintf("%06d",    as.numeric(PaidDateMonth)), PaidDateMonth = paste0(PaidDateMonth, "01"),  
      # Appending day to create YYYYMMDD format as string
      PaidDate = as.Date(PaidDateMonth, format = "%Y%m%d"),  # Convert to Date object
      Year = format(PaidDate, "%Y"),  # Extract Year as string
      # Create Season column based on PaidDate
      Season = case_when(
        format(PaidDate, "%m") %in% c("12", "01", "02") ~ "Winter",
        format(PaidDate, "%m") %in% c("03", "04", "05") ~ "Spring",
        format(PaidDate, "%m") %in% c("06", "07", "08") ~ "Summer",
        format(PaidDate, "%m") %in% c("09", "10", "11") ~ "Autumn",
        TRUE ~ "Unknown") ) %>%
    # Recode health boards to be more readable 
    mutate(HealthBoard = recode(HBT2014,
      "S08000032" = "Lanarkshire",
      "S08000031" = "Greater Glasgow & Clyde",
      "S08000024" = "Lothian",
      "S08000023" = "Lanarkshire",
      "S08000021" = "Greater Glasgow & Clyde")) %>%
    select(-HBT2014)  # Drop the original HBT2014 column
    return(full_prescribed_data)}
# 2019 prescription data
pitc_list <- list.files(path = "C:/Users/wurao/OneDrive/Documents/data_science/B205492/docs", 
         pattern = "pitc2019.*\\.csv$", full.names =TRUE)
health_boards <- c("S08000032", "S08000031", "S08000024")
# Process the data using the updated function
processed_data_2019 <- process_prescription_data(pitc_list, health_boards)
# Summarize total prescriptions by Health Board, Season, Year
seasonal_data <- processed_data_2019 %>%
  group_by(HealthBoard, Season, Year, ) %>%
  summarise( "Total prescriptions" = sum(PaidQuantity, na.rm = TRUE), .groups = 'drop')

2019 Trend

Figure 1

seasonal_data <- seasonal_data %>%
 mutate(HealthBoard = str_trim(HealthBoard)) #to remove any leading or trailing whitespace characters from the HealthBoard column in the dataset
seasonal_plot <- seasonal_data %>%
  ggplot(aes(x = Season, y = `Total prescriptions`, group = HealthBoard, color = HealthBoard)) +
  geom_line(linewidth = 1.2) +  # Line plot with specified line width
  geom_point(size = 3) +  # Add points to the plot
  labs(title = "2019 Seasonal Patterns of SSRI Antidepressant Use Across Health Boards", x = "Season",y = "Total Prescriptions" ) + 
  theme_minimal() +  theme(axis.text.x = element_text(angle = 45, hjust = 1))  # Rotate x-axis labels
# Print the plot
print(seasonal_plot)

The visualization of SSRI prescriptions across Scotland’s three major health boards in 2019 suggests a notable seasonal trend. Prescription rates appear to peak in the autumn months across NHS Greater Glasgow & Clyde, NHS Lanarkshire, and NHS Lothian, followed by a noticeable decline in winter, with spring and summer showing moderate levels. although this was unexpected this pattern implies a possible cycle where prescription demand rises as autumn approaches and diminishes in winter. To understand if this trend is consistent, further analysis will involve expanding the dataset to cover the 3 years prior. By examining a longer period,it will highlight whether these seasonal fluctuations are stable over time or influenced by other factor.

2016-2018 Trend

# Step 1: Listing Relevant CSV Files
pitc_list <- list.files(path = here("docs"), 
  pattern = "pitc(2016|2017|2018)(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\\.csv$", 
  full.names = TRUE) 

if (length(pitc_list) == 0) { stop("No matching files found. Check the file names or pattern.")}
# Step 2: Apply the function to process the files
health_boards <- c("S08000023", "S08000021", "S08000024")  # Customize this as needed
three_years_data <- process_prescription_data(pitc_list, health_boards)

Figure 2

# Summarize Data
Bar_prescribed_data <- three_years_data %>%
  group_by(Season, Year, HealthBoard,) %>%
  summarize(`Total prescriptions` = sum(PaidQuantity, na.rm = TRUE),       .groups = "drop")
# Bar Plot with ggplot2
p <- ggplot(Bar_prescribed_data, aes(x = Season, y = `Total prescriptions`, fill = Season, text = `Total prescriptions`)) +
  geom_bar(stat = "identity", position = position_dodge(width = 5), width = 0.9) +  # Adjusting spacing and bar width
  facet_grid(
    rows = vars(HealthBoard),  # Using rows for HealthBoard
    cols = vars(Year),
    scales = "free_y",  # Adjust y-scales independently
    labeller = label_wrap_gen(width = 15)  # Wrap long facet labels
  ) + labs(
    title = "Seasonal Patterns of SSRI Prescriptions by Year and Health Board",
    x = "Season",
    y = "Total Prescriptions"
  ) + theme_minimal() +
  theme(axis.text.x = element_text(angle = 45, hjust = 1),   # Rotating x-axis labels
        strip.text = element_text(face = "bold", size = 10, hjust = 0.5),  # Center facet labels
        plot.margin = margin(30, 50, 30, 30),  # Increase right margin for more space
        legend.position = "none",
        plot.title = element_text(hjust = 0.5, size = 10, face = "bold")  # Center the plot title
  ) + scale_fill_manual(values = c("Winter" = "blue", "Spring" = "green", "Summer" = "orange", "Autumn" = "brown")) +
  coord_cartesian(clip = "off")  # Prevent clipping of text outside plot area
# Convert ggplot to Interactive Plotly Plot
interactive_plot <- ggplotly(p, tooltip = c("x", "y", "text"))

# Show the Interactive Plot
interactive_plot

In Figure 2, data from 2016 to 2018 show consistent seasonal trends in SSRI prescriptions across Greater Glasgow & Clyde, Lanarkshire, and Lothian. In 2016, Autumn had the highest prescriptions: 8.88 million in Greater Glasgow & Clyde, 5.05 million in Lanarkshire, and 5.28 million in Lothian, with Winter consistently the lowest. Similar patterns were observed in 2017 and 2018, with Autumn prescriptions rising to 9.71 million in Greater Glasgow & Clyde, 5.61 million in Lanarkshire, and 5.83 million in Lothian by 2018.

Autumn consistently showed the highest prescription totals, followed by Spring and Summer, while Winter remained the lowest. These stable patterns across all health boards reinforce a seasonal cycle in prescribing habits, further explored through monthly data for more granular insights.

Figure 3

# Summarize the data and prepare for plotting
monthly_data <- three_years_data %>%
  group_by(Year, HealthBoard, PaidDateMonth,Season) %>%  # Ensure columns are correctly grouped
  summarize(`Total prescriptions` = sum(PaidQuantity, na.rm = TRUE), .groups = "drop") %>%
  # Create a Month_Factor to order months starting with March
  mutate(Month_Factor = factor(format(as.Date(PaidDateMonth, format = "%Y%m%d"), "%b"),  # Extract abbreviated month name from the date
    levels = c("Mar", "Apr", "May", "Jun", "Jul", "Aug", "Sep", "Oct", "Nov", "Dec", "Jan", "Feb")))
# Plot the monthly trends with seasons
ggplot(monthly_data, aes(x = Month_Factor, y = `Total prescriptions`, group = Season, color = Season)) +
  geom_line(size = 1) +  # Line plot with color for Season
  geom_point(size = 2) +  # Adding points for better visibility
  facet_grid(HealthBoard ~ Year, labeller = labeller(HealthBoard = label_wrap_gen(15))  # Wrap labels for HealthBoard
  ) + labs( title = "Monthly Trends in SSRI Prescriptions by Health Board, Year, and Season", x = "Month", y = "Total Prescriptions",color = "Season" ) + scale_color_manual(values = c("Winter" = "blue", "Spring" = "green","Summer" = "orange", "Autumn" = "brown")) +  # Custom colors for seasons
  theme_minimal() + theme( axis.text.x = element_text(angle = 45, hjust = 1),  # Rotate x-axis labels
    strip.text = element_text(face = "bold", size = 10), 
    legend.position = "bottom",  # Legend below the plot
    plot.title = element_text(hjust = 0.5),  # Center the plot title
    strip.text.y = element_text(margin = margin(r = 10)),  # Add margin to HealthBoard labels
    plot.margin = margin(t = 10, r = 10, b = 10, l = 20)  # Increase overall plot margins
  )

Observation of figure 3 shows that across the three years, autumn into the start of winter, particularly October, November, and December, consistently showed the highest SSRI prescription totals for most of Scotland’s major health boards. For Greater Glasgow and Clyde, October was the highest in 2018, followed by November in 2017. In Lothian, December was the peak month across all three years, and for Lanarkshire, November was the highest for 2017 and 2016, with August being the peak month for 2018. This pattern suggests that the autumn and early winter months—October, November, and December—tend to see higher SSRI prescriptions. This details could not be seen in the previous plots due to the aggregation of the months. However, observing the individual months provide insight into the finer details within each year.

Table

# Combine 2016-2018 and 2019 datasets
combined_prescribed_data <- bind_rows(Bar_prescribed_data, seasonal_data)
# Create the gt table
gt_table <- combined_prescribed_data %>%
  arrange(Year, HealthBoard, Season) %>%
  gt(rowname_col = "Season", groupname_col = "Year") %>%
  tab_header(
    title = md("**Seasonal Trends in SSRI Prescriptions Across Major Scottish Health Boards (2016-2019)**"),
    subtitle = "Overview of Seasonal Prescription Totals by Health Board" )%>%
  cols_label(
    Year = "Year",
    HealthBoard = "Health Board",
    Season = "Season",
   `Total prescriptions` = "Total Prescriptions") %>%
  fmt_number(columns = `Total prescriptions`,decimals = 0  # No decimals for prescription counts
  ) %>%
  tab_style(style = cell_text(weight = "bold"),locations = cells_row_groups(groups = TRUE)  # Bold year group labels
  ) %>%
  tab_options(
    row_group.as_column = TRUE,
    table.font.size = 14,  # Larger font size for readability
    data_row.padding = px(10),
    column_labels.font.size = 16,  # Bigger column headers
    heading.border.bottom.color = "gray85",
    heading.border.bottom.width = px(2)) %>%
  tab_spanner(
    label = "Health Board and Season",
    columns = c("HealthBoard", "Season")) %>%
  tab_style(
    style = cell_fill(color = "lightgray"),
    locations = cells_body(rows = seq(1, nrow(combined_prescribed_data), by = 2))  # Even rows   
    ) %>%
  tab_style( style = cell_fill(color = "lightblue"),
    locations = cells_body(rows = seq(2, nrow(combined_prescribed_data), by = 2))  # Odd rows
  ) %>%
  tab_style(style = cell_text(align = "center"),
    locations = cells_body()  # Center align body cells
  ) %>%
  tab_style(style = cell_text(align = "center"),
    locations = cells_column_labels()  # Center align column headers
  ) %>%
  tab_style(style = cell_text(weight = "bold", color = "darkblue"),
    locations = cells_column_labels(columns = c("Year", "HealthBoard"))) %>%
  tab_style(style = cell_text(weight = "normal", color = "black"),
    locations = cells_body(columns = c("Year", "HealthBoard", "Season"))) %>%
  tab_options(table.width = pct(100),  # Set the table to take up 100% of the width
    table.align = "center"   # Align the table to the center
  )

# Display the formatted table
gt_table
Seasonal Trends in SSRI Prescriptions Across Major Scottish Health Boards (2016-2019)
Overview of Seasonal Prescription Totals by Health Board
Health Board and Season
 Total Prescriptions
Health Board
2016 Autumn Greater Glasgow & Clyde 8,884,463
Spring Greater Glasgow & Clyde 8,821,739
Summer Greater Glasgow & Clyde 8,940,972
Winter Greater Glasgow & Clyde 8,594,210
Autumn Lanarkshire 5,052,455
Spring Lanarkshire 5,042,549
Summer Lanarkshire 5,104,842
Winter Lanarkshire 4,966,728
Autumn Lothian 5,280,164
Spring Lothian 5,249,669
Summer Lothian 5,296,597
Winter Lothian 5,121,263
2017 Autumn Greater Glasgow & Clyde 9,272,569
Spring Greater Glasgow & Clyde 9,212,639
Summer Greater Glasgow & Clyde 9,285,008
Winter Greater Glasgow & Clyde 8,941,147
Autumn Lanarkshire 5,317,885
Spring Lanarkshire 5,242,997
Summer Lanarkshire 5,330,759
Winter Lanarkshire 5,125,656
Autumn Lothian 5,529,307
Spring Lothian 5,528,153
Summer Lothian 5,501,288
Winter Lothian 5,382,146
2018 Autumn Greater Glasgow & Clyde 9,706,148
Spring Greater Glasgow & Clyde 9,537,012
Summer Greater Glasgow & Clyde 9,729,360
Winter Greater Glasgow & Clyde 9,391,876
Autumn Lanarkshire 5,612,085
Spring Lanarkshire 5,444,050
Summer Lanarkshire 5,629,655
Winter Lanarkshire 5,376,101
Autumn Lothian 5,832,617
Spring Lothian 5,755,843
Summer Lothian 5,768,094
Winter Lothian 5,670,963
2019 Autumn Greater Glasgow & Clyde 10,234,142
Spring Greater Glasgow & Clyde 10,062,044
Summer Greater Glasgow & Clyde 10,189,077
Winter Greater Glasgow & Clyde 6,630,138
Autumn Lanarkshire 5,931,818
Spring Lanarkshire 5,840,297
Summer Lanarkshire 5,946,353
Winter Lanarkshire 3,863,237
Autumn Lothian 6,210,689
Spring Lothian 6,141,095
Summer Lothian 6,104,339
Winter Lothian 6,017,089

Discussions

This study analysed seasonal patterns in SSRI prescriptions across Scotland’s three major health boards, revealing trends consistent with previous research conducted in northern countries. Prescription rates exhibited clear seasonal variation, with higher rates observed during autumn into the start of winter, which may be influenced by environmental and behavioural factors.

Comparison with Existing Literature

The findings in this report align with research by Harmatz et al. (2000) in the United States, which observed increased antidepressant use during autumn and winter, correlating with seasonal affective disorder (SAD). Reduced sunlight during winter months can exacerbate depressive symptoms, as sunlight plays a significant role in regulating circadian rhythms and serotonin levels. Scotland’s high latitude, characterized by particularly short winter days, may intensify this seasonal effect. Contrastingly, Sato et al. (2013) found less pronounced seasonal trends in antidepressant use in Japan. This disparity may reflect cultural differences in mental health stigma, healthcare access, and prescribing practices, as well as Japan’s milder seasonal variations compared to Scotland. Such findings underscore the complex interplay of environmental, social, and healthcare system factors that shape prescription trends.

Limitations and Future Directions

This study provides valuable insights into SSRI prescription patterns but is subject to several limitations. The analysis focuses solely on SSRIs, excluding other antidepressant classes that might exhibit different seasonal trends. Moreover, the lack of individual-level diagnostic information limits the ability to directly associate prescriptions with specific conditions such as SAD. For future direction research would benefit from addressing these limitations by incorporating demographic data, examining the usage patterns of non-SSRI antidepressants. Expanding the study’s geographic scope to include rural and remote areas which could provide a more comprehensive understanding of the role of healthcare access in seasonal variations.

conclusion

This study contributes to the growing body of evidence that seasonal factors influence SSRI prescribing patterns, particularly in high-latitude regions like Scotland. By exploring these trends, policymakers and healthcare providers can better understand the drivers of mental health treatment demand and implement strategies to improve the accessibility and timing of mental health care.

References

  1. Harmatz, M. G., Well, A. D., Overtree, C. E., et al. (2000). Seasonal variation in depressive symptoms in a general population. American Journal of Psychiatry.
  2. Madsen, T., Eplov, L. F., Mortensen, P. B., et al. (2015). Trends in antidepressant use: A nationwide study. BMJ Open.
  3. Lam, R. W., Levitt, A. J., Levitan, R. D., et al. (2006). Seasonal affective disorder: A review of the syndrome and its treatment. Canadian Medical Association Journal.
  4. Sato, T., Demura, S., & Kobayashi, Y. (2013). Seasonal differences in psychological and physiological changes in response to mental stress in a population without pronounced seasons. Psychiatry and Clinical Neurosciences.
  5. Jack, R. H., Joseph, R. M., Hollis, C., Hippisley-Cox, J., Butler, D., Waldram, D., & Coupland, C. (2023). Seasonal trends in antidepressant prescribing, depression, anxiety and self-harm in adolescents and young adults: an open cohort study using English primary care data. BMJ Mental Health, 26(1). https://doi.org/10.1136/bmjment-2023-300855

how AI helped me?

Function Development: Used AI to assist/polish in creating an R function for automating seasonal data filtering and summarization, streamlining the analysis process across multiple datasets. It also helped with struggle i had with my visualization such as giving ideas such as:Wrapping long facet labels when nothing was helping my labels not get cut off and recommending coord_cartesian.